htfa.groups.group_tree

htfa.groups.group_tree(groups, n_subjects, subject_ids=None)

Build the template tree of a grouped HTFA fit.

Parameters:

groupssequence of labels, sequence of sequences of labels, or GroupTree

One label (str or int) per subject, or a list of such sequences, one per level, top level first. For example, [site, subgroup] can group participants first by acquisition site and then by a subgroup within that site; these names illustrate nesting, not required labels. A tree from GroupTree.from_paths also supports unequal path depths and repeated folder names under distinct parents. For label sequences, groups must be nested: every label of a level belongs to one label of the level above (use distinct labels such as 'A-patient' and 'B-patient' for groups under different parents).

n_subjectsint

Number of subjects.

subject_idssequence, optional

One unique identifier per subject (per entry of X). Repeated identifiers raise ValueError: repeated runs are not independent subjects and must be combined before fitting.

Returns:

GroupTree

Raises:

ValueError

On labels that are not strings or integers, a level with the wrong number of labels, a label under two parents (crossed levels), or a subject identifier in two groups (crossed designs, such as the same people before and after a treatment, are not supported).

Examples

An illustrative age-group tree for four independent participants, with one global template above separate child and adult templates:

>>> tree = group_tree(["child", "adult", "child", "adult"], 4)
>>> tree.paths
[(), ('child',), ('adult',)]
>>> tree.solve(np.array([1.0, 4.0, 3.0, 6.0]))  # global, then group means
array([3.5, 2.5, 4.5])